Research map / Point clouds / Segmentation and detection
HPNet: Deep Primitive Segmentation Using Hybrid Representations
Siming Yan, Zhenpei Yang, Chongyang Ma, Haibin Huang, Etienne Vouga, Qixing Huang
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the descriptor module output (ablation) | HPNet · ANSI / ABCParts | Seg-IoU: drop by 14.9% and 17.5% | Improved |
| add the descriptor module output (ablation) | HPNet · ANSI / ABCParts | Type-IoU: drop by 6.5% and 13.4% | Improved |
| add the smoothness module (ablation) | HPNet · ANSI / ABCParts | Seg-IoU: 0.67% and 1.34% decrements | Improved |
| add the smoothness module (ablation) | HPNet · ANSI / ABCParts | Type-IoU: 0.17% and 1.13% decrements | Improved |
| add the consistency module (ablation) | HPNet · ANSI / ABCParts | Seg-IoU: 1.31% and 1.69% decrements | Improved |
| add the consistency module (ablation) | HPNet · ANSI / ABCParts | Type-IoU: 0.37% and 2.18% decrements | Improved |
| learn combination weights instead of fixed (ablation) | HPNet · ANSI / ABCParts | Seg-IoU: 0.08% and 0.46% decrements | Improved |
| learn combination weights instead of fixed (ablation) | HPNet · ANSI | Type-IoU: 0.00% decrements | No clear effect |
| combine both spectral descriptors (ablation) | HPNet (p+n) · ANSI | Seg-IoU: 92.41 -> 94.15 | Improved |
| combine both spectral descriptors (ablation) | HPNet (p+n) · ABCParts | Seg-IoU: 83.20 -> 85.24 | Improved |
| hybrid representations vs ParseNet | HPNet (p+n) · ANSI | Seg-IoU: 88.57 -> 94.15 | Improved |
| hybrid representations vs ParseNet | HPNet (p+n) · ABCParts | Seg-IoU: 82.14 -> 85.24 | Improved |
Later papers that compared with this method
3 later accepted papers report HPNet as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Constraint-Aware Feature Learning for Parametric Point Cloud (ICCV 2025 · better on 7 of 7 shared results)
- SpelsNet: Surface Primitive Elements Segmentation by B-Rep Graph Structure Supervision (NeurIPS 2024 · better on 0 of 1 shared result)
- Finding Good Configurations of Planar Primitives in Unorganized Point Clouds (CVPR 2022 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PrimitiveNet: Primitive Instance Segmentation With Local Primitive Embedding Under Adversarial Metric (ICCV 2021 · similarity 0.89)
- PartNet: A Recursive Part Decomposition Network for Fine-Grained and Hierarchical Shape Segmentation (CVPR 2019 · similarity 0.86)
- Deep FusionNet for Point Cloud Semantic Segmentation (ECCV 2020 · similarity 0.86)
- H3DNet: 3D Object Detection Using Hybrid Geometric Primitives (ECCV 2020 · similarity 0.85)
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.84)
- ODIN: A Single Model for 2D and 3D Segmentation (CVPR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
